We have realized on my team that the current ability for non service principal users to automate notebook runs through pipelines is not sustainable. There will be random failures due to authentication that happen after password changes are happen as a part of the normal semi-monthly cadence as required by most IT departments.
And when I say these failures are random I mean truly random. I have had 5 or so contributors making and automating notebooks with pipelines and only one of them has had this issue. But it happened on multiple pipelines and all on the same day about 4 days after he changed his password.
It can be resolved by making and saving any arbitrary update to the pipeline, but having these failures randomly start is a huge issue. Retries on the notebooks are ineffective.
I have already spoken with support and they have no obligation to raise this issue that is clearly a bug any higher because it is possible to solve the problem whenever it comes up just by re-saving the pipeline. But it is a huge issue that this comes up at all.
If this goes unresolved, all Fabric users should effectively expect that all of our pipelines may randomly start failing for no reason and that we need to babysit them all.
A more specific example of one of these errors is
Failed to get User Auth access token. The error message is: Failed to get User Auth access token. The error message is: AADSTS700003: Device object was not found in the tenant
